Why “Free RAM” is actually a waste of money 🐧💻

It’s common to run “free -m” on a Linux server and interpret a near-zero Free column as a sign of memory exhaustion.

In most cases, that is expected behavior rather than a problem.

Linux tries to keep RAM utilized because unused memory does not contribute to performance.

When memory is not being used by applications, the kernel uses it for the page cache. The page cache stores recently accessed file data so subsequent reads can be served directly from memory instead of going through the storage stack.

Since memory access latency is orders of magnitude lower than disk I/O (especially compared to HDDs and even NVMe SSDs), caching file pages in RAM significantly improves overall system throughput.

This is why the Free column in free often stays small on a healthy system. A more useful metric is Available, which estimates how much memory the kernel can provide to new processes without causing swapping. It includes reclaimable page cache and other reclaimable kernel memory.

Internally, the kernel manages cached pages using structures like XArray (introduced in Linux kernel 4.20), which replaced the earlier radix tree implementation.

XArray maps file offsets to cached pages and supports efficient concurrent lookups through mechanisms like RCU (Read-Copy-Update), allowing many CPUs to access the page cache with minimal locking.

In practice, a system with low “free” memory may simply be a system where RAM is being effectively used to cache data and reduce disk I/O.

hashtag#Linux hashtag#LinuxKernel hashtag#MemoryManagement hashtag#SystemsEngineering hashtag#KernelInternals hashtag#PerformanceEngineering hashtag#DevOps hashtag#SRE